UNIVERSITÀ DEGLI STUDI DI TRIESTE
XXXII
CICLO DEL DOTTORATO DI RICERCA IN
NANOTECNOLOGIE
PO FRIULI VENEZIA GIULIA - FONDO SOCIALE EUROPEO 2014/2020
VO
2
A PROTOTYPICAL
P
HASE
C
HANGE
M
ATERIAL
:
SPECTROSCOPIC STUDY OF THE ORBITAL
CONTRIBUTION ACROSS THE
M
ETAL
I
NSULATOR
T
RANSITION
Settore scientifico-disciplinare:
FIS/03
DOTTORANDO
A
LESSANDROD’E
LIACOORDINATORE
P
ROF.
A
LBERTOM
ORGANTESUPERVISORE DI TESI
P
ROF.
A
LBERTOM
ORGANTECO-SUPERVISORE DI TESI
D
R.
M
ARCELLOC
ORENO________________________________________________________________
“If, in some cataclysm, all of scientific knowledge were to be destroyed, and only one
sentence passed on to the next generation of creatures, what statement would contain
the most information in the fewest words?
I believe it is the atomic hypothesis (or the atomic fact, or whatever you wish to call it)that all things are made of atoms - little
particles that move around in perpetual motion, attracting each other when they are a
little distance apart, but repelling upon being squeezed into one another. In that one
sentence, you will see, there is an enormous amount of information about the world,
if just a little imagination and thinking are applied.”
R. Feynman
“
Please remember the deadline of dd/mm/yy for submitting the quarterly report. Let me remind you that its submission is mandatory.Failing to do so, you will no longer be
entitled to the funding and will have to return the scholarship received so
far.”
Index
Abstract ... 1
Chapter 1. ... 2
Vanadium oxides and VO2 ... 2
1.1 The Metal-Insulator Transition theory ... 7
1.1.1 Mott-Hubbard transition ... 7
1.1.2 Peierls transition ... 9
1.1.3 Anderson localization ... 10
1.2 VO2 structural properties ... 12
1.3 VO2 Band structure ... 14
1.3.1 Crystal Field theory ... 14
1.3.2 Ligand Field Theory ... 15
Chapter 2. ... 18
VO2 thin film and nanostructured film synthesis ... 18
2.1 Thin and ultrathin strained samples synthesis: MBE ... 19
2.1.1 Strain influence on structural properties ... 21
2.2 Nanostructured VOx samples synthesis: Supersonic Cluster Beam Deposition ... 23
2.2.1 Supersonic expansion of cluster beam ... 23
2.2.2 Pulsed Micro-Plasma Cluster Source ... 25
2.2.3 Stoichiometry determination: XPS ... 29
2.2.4 3d occupancy investigation: Auger L3M23M45 spectroscopy ... 30
2.2.5 Valence band study... 32
2.2.6 Vanadium oxides Work Function ... 36
2.2.7 Structural characterization: V L2,3 and O K edge XANES ... 39
Chapter 3. ... 45
Electronic structure investigation of VO2/TiO2 thin films and VO2 nanostructured films: Auger yield and ResPES ... 45
3.1 Auger Electron yield X-ray Absorption Spectroscopy ... 46
3.1.1 Auger Yield O KL23L23 ... 47
3.1.2 Auger yield V L3M23M45 ... 52
3.2 Resonant photoemission ... 57
Chapter 4. ... 66
Orbital contribution to the MIT studied by CIS spectroscopy. ... 66
4.1 V 3d CIS spectroscopy of VO2 thin films. ... 68
1
Abstract
VO2 is a fascinating 3d1 system undergoing a temperature triggered (67 °C) Metal Insulator
Transition (MIT) coupled with a structural phase transition, from a low-temperature monoclinic insulator to a high-temperature tetragonal metal. Since its discovery, the MIT has been widely studied with a twofold interest: its applicative potential and its nature.
Different theoretical models have been proposed to explain the occurrence of the insulating phase of VO2 like a structurally driven Peierls transition or a Mott-Hubbard transition triggered
by electron mutual Coulomb repulsion. However, a clear theoretical picture is missing since VO2
properties are determined by a complex interplay among lattice, orbital and electronic degrees of freedom. Therefore, in order to exploit the MIT features for technological application, a detailed study of the influence and interplay between the different degrees of freedom is of paramount importance.
With the aim of disentangling the lattice-orbital-electronic intrigue, in this thesis, four samples with different structural properties have been studied. Three thin strained films and one nanostructured disordered VO2 film have been investigated using advanced spectroscopic
2
Chapter 1.
3
In the last decades, technological advancements in the field of information processing, and storage have revolutionized our lifestyle. The advent of high performance computing, instantaneous worldwide range communication as well as fast storage of large dataset had irrevocably changed society. Computer assisted design of new products and simulations to predict functionalities, the ability to manage complex tasks like control of robotic manufacturing processes or handle supply chains on the global scale are just a few examples of the influence of the technology on our everyday life.
The key element of this technological advancement is silicon, which plays a major role in semiconductor-based electronics. The success of silicon is due to three main features: 1) its low cost and high abundance, 2) the modulation of flow of electrical charges applying an electric field, enabling storage, manipulation and amplification of electrical signals, 3) the conversion of optical into electric signal and vice versa.
The improvement of fabrication technologies led to the continuous miniaturization of silicon-based electronics and in particular of transistors in agreement with the Moore law[1]. The prediction to double the number of transistors per unit area every two years has been respected for at least 40 years, although a slow down has been observed since 2015 due to the increasing complexity of the miniaturization procedures and chips manufacturing costs[2].
The decreasing trend of the Moore law points out that the margin of the silicon based technologies improvement is vanishing, triggering the R&D of materials with novel properties and functionalities. In this framework, Transition Metal Oxides (TMO) are possible candidates. In ordinary semiconductors, the valence band is generated by the superposition of s-p orbitals while in TMO it is formed by d electrons, which provide new functional properties. Usually, in TMO there is a complex interplay between structural, orbital and electronic degrees of freedom, which give rise to unconventional and fascinating properties: the appearance of high-temperature superconductivity in complex cuprates[3] and the observation of giant magnetoresistance in manganites[4] are just few examples.
In contrast to middle and late TMOs, early TMOs are far less studied and in particular vanadium oxides disclose a huge applicative potential[5].
Vanadium ([Ar] 3d34s2) is a very reactive element and different stoichiometric oxides (VO, V 2O3,
VO2, and V2O5) characterized by different oxidation states such as V+2, V+3, V+4 and V+5 as well as
mixed-valence oxides (Magneli and Wadsley series)[6] can be synthesized. This huge variety of oxides implies an equally large variety of structural and electronic properties, which can have different and powerful technological applications. As an example, because of their layered structure, both V2O5 and V6O13 have been deeply investigated for possible applications as
cathode materials in Li-ion batteries[7–9]. More importantly, this class of oxides bears the seed of a strong electronic correlation. VO2, V2O3 and the mixed-valence VnO2n-1 (n=3-6, 8, 9) systems
all exhibit first-order metal-insulator transitions (MIT)[5,10].
One of the most studied oxides, the tetravalent oxide VO2, undergoes a sharp, reversible,
hysteretic, MIT coupled to a structural phase transition, passing from a high-temperature tetragonal metal to a low-temperature monoclinic insulator phase.
Vanadium dioxide MIT can be triggered by temperature (~67 °C in single crystals)[11], electric field[12] or photo-carrier doping[13], making VO2 an extremely versatile and attractive
4
amplitude (ΔA), width (ΔW), sharpness (ΔH) and critical temperature (Tc), which are dependent
on the VO2 structure at micro and nanoscale.
Figure 1.1: Qualitative representation of the hysteresis cycle of the resistance across the phase transition with the characteristics parameters that define the MIT features. ΔA is the amplitude of the transition, ΔW is the width and
ΔH the sharpness or slope.
Narayan and Bhosle proposed a qualitative model based on the classical nucleation theory[14] which correlates ΔA, ΔW and ΔH (see Fig. 1.1) to samples microstructure being also able to rationalize results available in the literature[15–17]. They rely on the characterization of the phase transition in terms of the Gibbs free energy variation in adiabatic approximation:
∆𝐺 = ∆𝑇∆𝑆 (1.1) where ΔG, ΔT and ΔS are the variations of the free energy, the temperature and the entropy across the phase transition. According to the classical nucleation theory, the minimum radius necessary to an agglomerate to start the nucleation process (i.e., the critical radius) can be written as:
𝑟𝑐 = 2𝛺
∆𝐺 (1.2)
in which Ω is the interfacial energy. The main features of the Narayan-Bhosle semi-empirical model are listed in the following:
1. The sharpness of the transitionΔH is directly dependent on the defect density per unit of volume, including point defects, cluster impurities, and grain boundaries ∆𝐻 = 𝐶𝜌𝑑 (1.3)
where C is a constant and ρd the density of defects. Usually, grain boundaries defects are
not relevant respect to point defects, until the grain size decreases to the nanometer dimension. In this case, they will play a role in decreasing the sharpness of the transition[18].
2. ΔA is inversely proportional to defect density. High purity VO2 exhibits a larger transition
5
3. The finite temperature ΔT necessary for the phase transition is directly proportional to the hysteresis width ΔW. Therefore, a minimum hysteresis will always be present even in the
purest sample.
∆𝑊 ∝ ∆𝑇 =∆𝑆𝑟2𝛺
𝑐 (1.4)
Reducing the grain radius and increasing the interfacial energy, the hysteresis width increases. Therefore both sample morphology and dimension affect ΔW. Various studies on VO2 nanoparticles clearly showed the increase of the transition width reducing the
nanoparticle size[16,19,20]. The explanation of the ΔW dependence from the interfacial energy Ω can be found in grain boundaries. In synthesis, the interfacial energy is minimized when two adjacent grains are oriented within small angle respect to a fixed direction, hence the polycrystalline samples transition width is larger respect to single crystal[21].
Using this model is possible to predict the phase transition features for different complex morphologies. This is extremely important in order to synthesize VO2 samples with the necessary
properties to match the desired application. For example, for memory device applications a VO2
sample with a large hysteresis is preferable, while for sensing applications, the highest efficiency is required and therefore a sharp phase transition with a large amplitude is preferred. In general, from the Narayan-Bhosle semi-empirical model some general trends can be extrapolated. For high quality single crystals, i.e., big grains well oriented, the model predicts a sharp transition (small ΔH) with a large amplitude (ΔA) and a small width (small ΔW). Respect to single crystal, in randomly oriented polycrystalline films, the MIT will exhibit a smaller ΔA (because of the increased amount of defects) and a larger width caused by the mismatch of large grains orientation. In nanoparticles assembled films, the small grain size will contribute, in addition to the increase of the number of defects and to the random grain orientation, to the larger width and to smaller amplitude phase transition respect to polycrystalline films. In amorphous VO2
samples, the phase transition is broad (large ΔH) because of the high concentration of defects while the amplitude is reduced. The width ΔW is also reduced because of the reduction of the grain boundaries[22,23]. In addition, epitaxial strain[24] and alloying[25] allow to increase or decrease the transition critical temperature
In this scenario, the VO2 emerges as a prototypical PCM since its near room temperature phase
transition is easily accessible and easy to manipulate. Consequently, different applications based on VO2, MIT exploitation have been proposed in different technological areas.
In electronics, PCMs and in particular VO2 have been proposed for neuromorphic
computing[26,27], radiofrequency and electrical switches, augmented Field Effect Transistor (FET)[5] and coupled oscillators[28], though a physical model for the simulations of PCM based devices to calculate heat transport, Joule heating and MIT features has been addressed only recently[29]. The MIT in VO2 can be controlled at the nanoscale using an atomic force
microscope equipped with a biased conducting tip, suggesting applications in computing and memory devices[30].
VO2 thin films, used as a window coating, have also been proposed as an energy-saving material
with the ability to decrease the energy loss in buildings. Indeed, in the low-temperature insulating phase, VO2 is almost transparent to infrared (IR) radiation. Rising the temperature, in
6
W doping [31], it is possible to use VO2 coatings to modulate the infrared component of the
radiation entering or leaving a building, therefore, controlling its thermal isolation.
Despite the attention received for its possible applications, since the first observation of the VO2
phase transformation by Morin in 1959[11], the very nature of the MIT has been the argument of intense scientific debate. The abrupt jump in resistivity is accompanied at the same temperature by a crystalline transition, going from the low-temperature monoclinic phase to the high-temperature rutile structure. Different models have been proposed to explain the VO2
Metal to Insulator Transition (MIT): a strong electron correlation driven Mott-Hubbard transition[32–34], a Peierls structural distortion driven MIT[35–37] or a cooperative Mott-Peierls mechanism[38].
The understanding of the mechanism at the origin of MIT is crucial in order to improve the phase transition. However, this is complicated by the coexistence of different driving forces. Because of the complex interplay between charge, lattice and orbital degree of freedom, an unambiguous theoretical or experimental explanation of the MIT is not yet available. To experimentally approach this long-standing problem, it is necessary to characterize across the phase transition a coherent set of samples with controlled properties, using techniques, which concurrently probe more than one degree of freedom.
In this thesis, two sets of VO2 samples (three crystalline strained films and one disordered
nanostructured films) have been characterized and studied using different experimental techniques. Thin and ultra-thin crystalline samples have been characterized using X-ray Photoelectron Spectroscopy (XPS, Appendix 1), Resonant Photoemission (ResPES, appendix 3.1) and X-rays Absorption Near Edge Structure (XANES, Appendix 2) acquiring two different Auger Channels O KL23L23 and V L3M23M45. The nanostructured (NS) VO2 samples have been
characterized using XPS, Ultra Violet Photoelectron spectroscopy, Work function measurements, XANES in total electron yield and Transmission electron microscopy. The details of the samples’ preparation and characterization are described in Chapter 2.
7
1.1 The Metal-Insulator Transition theory
The phenomenological definition of a metal and an insulator is quite straightforward: a metal is a material capable to conduct electricity while an insulator does not. To have a more rigorous definition of the metallic and insulator state we had to wait until the early ’30s of 1900 when the band theory has been developed[39]. The first real breakthrough was the demonstration of the Bloch theorem, which explains the quantum-mechanical motions of electrons in a periodic lattice potential. The eigenvalues of the solution of the Schrödinger equation proposed by Bloch, are the energy bands shaped by the electron crystalline momentum k, and it can be seen as the solid-state analogue to the discrete energy levels in atoms. Within the band theory of solids, a metal is defined as a material with a partially filled valence band whereas an insulator exhibits a completely filled band. The classification of solid-state properties using band theory has been extremely successful. However, not all materials with a predicted partially filled band (in particular TMOs) show the conducting behaviour[40].
To understand this discrepancy, it is necessary to remember that band theory and the Bloch theorem have been developed within the independent particles approximation. As proposed by the Nobel prize N.F. Mott[41], introducing the Coulomb interaction among electrons we observe spatially localised electronic wave functions. Including a mutual electron repulsion, it is possible to explain the insulating behaviour of some oxides (e.g., NiO) and predict metals insulator transitions. However, electron correlation is not the only way in which an MIT can occur.
Figure 1.1.1: Three main mechanisms for a metal-insulator transition (MIT).
A complete theoretical description of the criteria that leads to MIT in solids is behind the purpose of this thesis and a rigorous treatment is available in the literature[42]. In the next sections, the three main MIT mechanisms will be described: the electron correlation induced Mott-Hubbard transition, the structurally driven Peierls transition and the disorder based Anderson localization.
1.1.1 Mott-Hubbard transition
8
Coulomb interaction is higher than the kinetic energy dispersion there is an energy gain if the d or f wave functions are localized around the metal ion instead of being itinerant i.e. a bandgap is opened.
These classes of compounds are called Mott insulators. Mott proposed a general criterion (Mott criterion) according to which a conventional band metal actually behaves as an insulator when: ahN1/3≲ 0.25 (1.1.1)
where ah is the effective Bohr radius of the valence electrons, and N is the electron density.
Mott suggested that the metallic state occurs in these materials as an excited state while the (Mott) insulator electronic structure is associated with electron-hole excitons[43]. An equivalent point of view is to consider that in strongly correlated systems the conduction happens through tunneling mechanism. Therefore, the presence of insulating or metallic states is determined by the relative magnitude of the mutual Coulomb repulsion between two electrons on the same site and the energy transfer integral.
9
1.1.2 Peierls transition
Peierls transition is a structural phase transition occurring in one-dimensional (1-D) systems. In an equally spaced atomic mono-dimensional system, the electron-phonon interaction is such to lead to the formation of atomic dimers doubling the lattice period. A bandgap at the FL is opened as a consequence of the change in periodicity, turning the system from conductive to insulating. Despite a pure 1-D system is an ideal model, Peierls transition is of extreme interest for materials, which possess 1-D-like features, e.g.: polymer chains, 3-D solids with a strongly directional orbital order or quasi 2-D layered materials, e.g., transition metal dichalcogenides, which exhibit a super-lattice distortion with the doubling of the lattice period.
Figure 1.1.2: Schematic representation of the Peierls distortion. Top panel: 1-D chain of monovalent atoms along with the density of charge distribution (ρ) and the energy-crystalline momentum dispersion relation in the undistorted scenario. Bottom panel: atomic dimerization with the consequent modulation of the charge distribution
and doubling of the lattice constant. In this distorted scenario, the bandgap opens at kF.
10
the periodic potential contributions it can be demonstrated that the minimum energy configuration is that corresponding to the dimerization of the atomic chain with the concurrent opening of the bandgap at kF. In the lower panel of Figure 1.1.2, is reported the distorted
configuration with the gap opening.
The pairs of atoms double the lattice period and create two different bond lengths. The opening of the bandgap in the single-particle excitation spectrum, push the electrons around kf toward
lower energy levels respect to the unperturbed case. The Peierls distortion is energetically favourable when the energy savings due to the band gaps opening outweighs the elastic energy necessary for ion rearrangements. In the distorted chain the electrons condensate forming electrons-holes pairs is characterized by a collective mode with wave vector 2kf and the charge
density distribution defined as [44]:
𝜌(𝑟) = 𝜌0+ 𝜌1cos (2𝑘𝑓𝑟 + 𝜑) (1.1.2)
where ρ0 is the unperturbed charge density (constant), ρ1 is the amplitude of the oscillation and
φ is the phase shift. The condensate is called Charge Density Wave and its formation is mainly
a 1-D phenomenon, even if it has been observed also in 2-D and 3-D systems [44]. It is important to underline that the Peierls transition is mainly observed when thermal excitations are minimized, i.e. at low temperature. Increasing the temperature, the atomic chain may acquire enough vibrational energy to arrest the dimerization and therefore to close the bandgap. The Peierls distortion can be experimentally observed using all those techniques able to track, directly or indirectly, the atomic positions within a solid. X-ray diffraction (XRD), X-ray Absorption Fine Structure (XAFS) for bulk materials and Low Energy Electron Diffraction (LEED) for 2-D materials can directly probe the atomic pair formation and the doubling of the unit cell across the transition. Techniques like Raman spectroscopy and XANES can be also used to probe the transition from the spectroscopic signature (the vibrational mode and the empty band formation, respectively) following the atoms dimerization.
In this thesis, XANES spectroscopy has been used to investigate the empty band evolution across the VO2 phase transition for samples with different structural properties. As will be explained in
section 1.3.2, the Peierls distortion results in the spectroscopic fingerprint of the empty density of states, which evolution has been studied by means of XANES.
1.1.3 Anderson localization
11 The main results of the Anderson analysis are two:
in the case of one and two dimensions solid, an arbitrary degree of disorder results in the localization of the wave functions;
in the case of three-dimensional solids exists a critical value of W above which the conduction is inhibited. W is the width of the stochastic distribution of the eigenvalues of the Anderson Hamiltonian.
W is usually as big as the mean Schrödinger equation eigenvalues energy, therefore a perturbative approach cannot be pursued. However, advanced numerical approaches are being developed in order to simulate and predict the disorder-induced degree of localization in solids. In the Anderson MIT, the FL occupation is coupled to the transition from disorder to order. The combination of experimental techniques sensitive to sample order and to FL occupation can be exploited to study the Anderson localization. In this thesis, we used the combination of XANES and ResPES to investigate the local order and the evolution of the metallicity of the VO2
12
1.2 VO
2structural properties
The crystalline structure of VO2 is different in the metallic and insulating states. In the
high-temperature metallic phase (above 67 °C in single crystal) VO2 has a lattice structure belonging
to the tetragonal symmetry. This phase is commonly referred to as rutile. The vanadium atoms are positioned at (0,0,0) and (½, ½, ½) with ar=br=4.554 Å and cr=2.851 Å [48,49]. According to
the Pauling rules[50], vanadium atoms are surrounded by oxygen octahedra, which can be considered the basic coordination complex of the crystal structure. The octahedron sites share edges only along cr (rutile c axis) which introduces a small orthorhombic distortion within the
octahedron[36] generating two different V-O distances: equatorial end apical. The two apical oxygen have position ± (u, u, 0) while the four equatorial oxygen atoms ±(±(u-½), ∓(u-½); ½)
where u=0.3001 at 360 K[48,49]. The equatorial distance is between the metal atom and the four neighboring oxygen atoms with z =zmetal ± ½.
Apical and equatorial V-O bond lengths depend on the unit cell edges length, i.e. lattice parameters. In particular apical V-O (JA) linearly depends on the ar length:
𝐽𝐴𝑝𝑖𝑐𝑎𝑙= 𝐽𝐴= √𝑢2𝑎𝑟2+ 𝑢2𝑏𝑟2= √2𝑢2𝑎𝑟2 = √2𝑢𝑎𝑟 (1.2.1)
The equatorial bond length (JE) includes all the lattice parameters.
𝐽𝐸𝑞𝑢𝑎𝑡𝑜𝑟𝑖𝑎𝑙 = 𝐽𝐸 = √(𝑢 − 1 2) 2𝑎 𝑟 2+ (𝑢 −1 2) 2𝑏 𝑟2+ 1 4𝑐𝑟 2= √2(𝑢 −1 2) 2𝑎 𝑟 2+1 4𝑐𝑟 2 (1.2.2)
Figure 1.2.1: Schematic representation of the high temperature tetragonal (left) and low temperature monoclinic (right) unit cell. Only V atoms are depicted for clarity (red dots). In the low-temperature monoclinic phase, the unitary cell (dashed black line) is doubled respect to the tetragonal structure (continuous black line). The V-V dimerization and tilting results in the appearance of two V-V distances. The values are taken from[51]. The doubling of the unit cell and the V-V dimerization are a fingerprint of a Peierls transition.
The low-temperature phase is the distorted high-temperature structure and belongs to the monoclinic symmetry. Across the phase transformation, the vanadium atoms pairs and tilt along the rutile c axis. As can be seen in Figure 1.2.1 the unit cell doubles the size respect to the tetragonal phase and two different V-V lengths appear along the am (cr) axis. The sum of the
13
transition. The crystal structure parameters are: am=5.751 Å, bm=4.537 Å, cm=5.382 Å,
β=122.646°[52].
This insulating phase is called M1. The structural phase transition is usually concomitant with the MIT. However the influence of external parameters like strain[53,54] or pressure[55], or studying the photo-induced transition on the ultra-fast timescale[13,56], MIT and Structural Phase Transition (SPT) can be separated, and different monoclinic distorted metastable phases can be observed e.g. M2 and T[57]. M2 phase is characterized by the existence of two sub-lattice
of vanadium atoms: in the first, they are paired along the cr axis but not tilted, in the second
they are not paired but tilted perpendicularly to the cr axis. The T phase is intermediate between
the M1 and M2 in which part of the vanadium atoms are unpaired and part are tilted.
The existence of the metastable insulating phases of VO2 can no longer be explained with the
14
1.3 VO
2Band structure
To have a deep insight into the MIT mechanism, it is necessary to understand how the band structure evolves when crossing the phase transition. Understanding how the orbital features are connected to the other degrees of freedom involved in the MIT is of crucial importance. In VO2 the band structure is the result of the respective arrangements of vanadium and oxygen
atoms, electron correlation and O-V hybridization. In TMO the oxygen atoms have a complex influence on the energy levels of the TM atom. The first attempt made to explain the effect of the oxygen (or ligand) geometric arrangement on the transition metal orbitals has been made by Bethe and Van Vleck in the early years of 1930 [58,59] with the introduction of crystal field theory.
1.3.1 Crystal Field theory
Crystal field theory (CFT) describes the degeneracy breaking of the outer valence orbitals (in the case of study V 3d orbitals) induced by an electrostatic negative charge distribution (anions or ligands). This theory has been often applied to describe TM coordination complexes, in which the central atom usually metallic (coordination centre) is surrounded by oxygen atoms or molecules with ligand character.
A brief overview of CFT highlighting its main characteristics is proposed in this section.
In CFT the coordination centre and the ligands are respectively considered as positive and negatives point charges. The theory defines the energy changes undergone by the outer 3d manifold of TM atoms when surrounded by oxygen atoms. The interaction between the ligand and coordination centre orbitals are not taken into account in this theory. The proximity of the ligand atoms to the coordination centre, implies that the ligand is closer to certain 3d orbitals and further to others, lifting the degeneracy. The details of the energy splitting are mostly dominated by:
The nature of the ligands and of the metal atoms,
The coordination of the central atoms, which determines the geometric arrangements of the ligands.
In vanadium oxides, the ligands dispose at the vertex of an octahedron with the vanadium atom in the middle [50]. The octahedral coordination is one of the most common in transition metals oxides; in addition, the other coordination geometries (tetrahedral, trigonal, etc.) can be described starting from the octahedral case. To understand how the degeneracy is removed by the octahedral crystal field, it is necessary to remember the orientation of the five 3d orbitals. Given a Cartesian frame of reference XYZ (see Figure 1.3.1a), the five d orbitals are directed toward different directions. The orbital 𝑑𝑧2 points toward the Z axes, 𝑑𝑥2−𝑦2 toward the X and
Y axes and 𝑑𝑥𝑦 𝑑𝑧𝑥 and 𝑑𝑦𝑧 are directed along the bisector of the XY, ZX and YZ planes
15
central atom, the atomic degenerate 3d manifold is lifted in energy since al the orbitals feel the same electric field.
Figure 1.3.1: a) Vanadium (red) and oxygen (blue) atoms arrangement. The frame of reference XYZ is also depicted for clarity. b) Diagram of the energy level splitting in an octahedral crystal field
If 6 negatively charged ligands are placed: 2 along the Z-axis and 4 along the XY plane bisectors (as in figure 1.3.1a) the 3d manifold is split as reported in Figure 1.3.1b. The orbitals 𝑑𝑧2 and 𝑑𝑥𝑦
which points directly along with the ligands, undergoes a higher repulsion by oxygen presence therefore they are lifted in energy. On the other hand, the orbital 𝑑𝑥2−𝑦2, 𝑑𝑧𝑥 and 𝑑𝑦𝑧 pointing
between the ligand atoms are less influenced. There is a loss of degeneracy that results in the splitting of the 3d manifold in two groups of orbitals: 𝑑𝑧2, 𝑑𝑥𝑦 and 𝑑𝑥2−𝑦2, 𝑑𝑧𝑥 and 𝑑𝑦𝑧.
According to group theory, the octahedral symmetry belongs to the point group Oh. 𝑑𝑧2, 𝑑𝑥𝑦
belong to the irreducible representation eg whereas 𝑑𝑥2−𝑦2, 𝑑𝑧𝑥 and 𝑑𝑦𝑧 to the t2g. The
octahedral field has the effect of splitting the 3d manifold in two high energy eg and three low
energy t2g levels. The energy separating the two groups of orbitals Δoct (or often 10Dq) is called
crystal field splitting. Referring to the energy of the spherical field, the octahedron geometry lifts the eg levels of 0.6Δoct while lower the t2g of 0.4 Δoct.
The crystal field splitting can be measured through absorption spectroscopies (UV-VIS or XANES) and it is a direct measure of the order of the geometric arrangement of the ligands around the central atom. Distortion effects of the Jahn-Teller type can additionally split the energy levels and thus reduce Δoct allowing to extract information about the short-range order of the sample.
1.3.2 Ligand Field Theory
Ligand field theory (LFT) is the application of molecular orbital theory to transition metal complexes. The basic of LFT is the introduction of the interaction between the metal and ligand orbitals within the CFT framework. It has been developed at the end of 1950 by B. S. Griffith and L. Orgel [60] with the purpose to increase the accuracy of CFT.
Orbitals combination between metal and ligands are allowed only between states with the same symmetry. In an octahedral field, the oxygen atoms form σ bonds with the 𝑑𝑧2 and 𝑑𝑥𝑦 orbitals.
In VO2 the 𝑑𝑧𝑥 and 𝑑𝑦𝑧 forms π type bonds with the oxygen orbitals leaving 𝑑𝑥2−𝑦2 non-bonding.
The first description of the VO2 band structure by LFT is due to J.B. Goodenough in 1971[36]. The
16
Figure 1.3.2: Oxygen-Vanadium orbital contribution to the VO2 (metallic phase) band structure. The V t2g orbitals are
further split because of the orthorhombic distortion into one non-degenerate a1g and twofold degenerate eg. The
grey-shaded orbitals are occupied by electrons, while the white orbitals indicate empty states. Only the V 3d and O 2p orbitals are depicted for clarity.
Crystal field-effect splits the degenerate 3d manifold into 3 t2g and 2 eσg levels. The 2 degenerate
eσ
g overlap with oxygen p orbitals with σ character (pσ). The bonding and anti-bonding
combinations of the eσ
g-pσ bonds are labelled as σ and σ*. A small orthorhombic distortion
directed along the cr axis (X-axis in the XYZ frame of reference of Figure 1.2.1a), further splits
the 3 t2g levels in one not degenerate a1g and twofold degenerate eπg states. The eπg orbitals
hybridize with the π symmetric O 2p orbitals (pπ) while the a1g orbital is not involved in the V-O
bonding. The eπ
g -pπ orbital admixture results in a bonding π band and in an anti-bonding π*
band. The a1g is formed by the 𝑑𝑥2−𝑦2 orbital and is directed along the cr (am) axis and is
responsible for vanadium-vanadium intermetallic bonds. The latter orbital is historically called 𝑑||.
The major spectral changes across the MIT occur at the Fermi Level (FL), in Figure 1.3.3 are depicted the detailed orbital evolution of the VO2 band structure crossing the phase
transformation.
In the metallic phase (Figure 1.3.3a) the 𝑑|| and π* orbitals cross the FL and are partially
degenerate. Their relative position (respect to the FL) and occupation depend on the hybridization between V 3d and O 2p orbitals and by the cr/ar ratio as will be explained in section
2.1.1. In bulk crystalline VO2 π* is less occupied (or more empty) respect to 𝑑||. For the
occurrence of MIT the 𝑑|| needs to be stabilized (lowered in energy) and π* destabilized (shifted
to higher energy) respect to the FL. The structural distortion described in the previous section, pairs and tilts V atoms moving them away from the center of the octahedron.
In the insulating phase (Figure 1.3.3b) the dimerization of vanadium atoms increases the overlap of the 𝑑|| orbital on different V sites, therefore, stabilizing to lower energy the 𝑑|| orbital while
creating the empty 𝑑||∗ (t2g character) band. The off-centre position of V atoms in the insulating
17
stabilization of the π band and therefore the consequent destabilization of the π*, which is shifted above the FL [36].
Figure 1.3.3: Band structure modifications across the MIT near the Fermi level. a) metallic phase, b) insulating phase. The grey-shaded orbitals are occupied by electrons, while the white orbitals indicate empty states.
In the insulating phase (Figure 1.3.3b) the dimerization of vanadium atoms increases the overlap of the 𝑑|| orbital on different V sites, therefore, stabilizing to lower energy the 𝑑|| orbital while
creating the empty 𝑑||∗ (t2g character) band. The off-centre position of V atoms in the insulating
state increases also the overlap between vanadium and oxygen atoms which results in increased stabilization of the π band and therefore the consequent destabilization of the π*, which is shifted above the FL [36]. The 𝑑||∗ is directed along the cr axis in the metallic phase and along the
V-V dimer in the insulating phase and it is strictly related with the unidimensional V-V dimer chain formation in the monoclinic insulating phase [61], while the π* is more isotropic within the lattice [62]. The model proposed by Goodenough depicts a structurally driven phase transition triggered by V-V pairing and tilting. However, Zylbersztejn and Mott argued that the vanadium atoms dimerization is not necessary to the opening of the bandgap since M2 and T
18
Chapter 2.
VO
2
thin film and
19
VO2 MIT features (transition temperature, hysteresis width and amplitude) are deeply
influenced by sample structure (single crystal, polycrystalline, disordered samples, etc.) [63]. In section 1.3.2 it has been shown the importance of vanadium and oxygen relative position to determine the electronic structure properties of VO2. V-O arrangements can be controlled
during the materials synthesis stage and, thus having the possibility to study the VO2 electronic
structure under different lattice conditions. The production of VO2 films with different structural
properties is therefore essential in order to understand the combined influence of electronic and lattice degrees of freedom on MIT.
Two main approaches have been adopted to modify vanadium and oxygen atoms disposition within the lattice: synthesis of strained samples and synthesis of NS samples.
In this chapter, a survey of the synthesis methods used to produce the samples discussed in this thesis is presented. Section 2.1 is devoted to a brief overview of Molecular Beam Epitaxy (MBE) technique, which allowed to synthesize strained crystalline samples. In Section 2.2 the use of Supersonic Cluster Beam Deposition (SCBD) method for NS VOx films production is presented.
This method allowed us to obtain disordered VO2 films.
2.1 Thin and ultrathin strained samples synthesis: MBE
Lattice strain is an effective way to slightly modify in a controlled fashion the atoms' positions within a solid. Strain application relies on the ability to synthesize thin crystalline samples on a properly oriented substrate. The lattice mismatch between the sample and the substrate must not be too severe therefore the substrate needs to have the same lattice symmetry of the sample deposited but with different unit cell edges length. In the early stage of growth, the lattice of the deposited sample undergoes a strain (tensile or compressive depending on the substrate choice) in order to match the lattice parameters of the substrate. Increasing the sample thickness the lattice relaxes to its bulk value.
Strain control has been proven as an effective method to tune MIT features (critical temperature, orbital occupancy, etc. ) in VO2 [62,64,65]. The interest in developing a technique
able to synthesize single-crystalline samples controlling their cell edges length is twofold: the MIT control and the understanding of its nature.
MBE matches all the latter requirements allowing the synthesis of samples of pure VO2 with
different degrees of strain.
Epitaxy refers to the discipline of growing crystalline materials on a crystalline substrate. Among the different deposition techniques developed to produce crystalline overlayers, MBE proved its importance in the synthesis of monocrystalline thin films of complex oxides [66].
MBE was developed by J. R. Arthur and Alfred Y. Cho in the late 60s at Bell laboratories [67,68] and it takes place in ultra-high-vacuum (UHV) conditions (10-10 mbar). In MBE the material of
20
Figure 2.1.1: schematic representation of a MBE deposition apparatus.
The strength point of MBE is the constant deposition rate (typically of the order of 0.1 Å/s) that allows the films to grow epitaxially. The amount of heat supplied to the crucible defines the rate of the metals evaporation and the deposition rate on the substrate.
In order to produce single-crystal metal oxide thin films, it is necessary to introduce oxygen in the experimental chamber using a mass flow controller.
Radio Frequency (RF) assisted Oxide MBE is often performed. In this variant of the MBE process, an RF plasma source is used to provide reactive oxygen radicals that will interact with the metals atom at the moment of deposition.
In this thesis we investigated films of VO2 with a thickness of 8, 16 and 32 nm deposited on a
clean substrate of TiO2 (001) by the RF-plasma assisted oxide-MBE instrument, working with a
base pressure better than 4x10-9 mbar. At a constant growth rate of 0.1 Å/s, the thickness was
controlled by adjusting the deposition time in a range from several unit cells to tens of nanometers. During the deposition, the substrate has been kept at 550 °C.
The interfacial cross-section has been investigated with a high-resolution scanning transmission electron microscope (STEM). High angle annular dark-field (HAADF) STEM images were taken on a JEM ARM200F with a probe aberration corrector, while the diffraction pattern was acquired on a JEM 2100 TEM. The complete details of the epitaxial film preparation and characterization performed at the University of Science and Technology (Hefei, P.R. China) are reported elsewhere [64,65].
The samples have been produced in China and moved to Elettra synchrotron radiation facility for ex-situ investigation.
21
2.1.1 Strain influence on structural properties
The TiO2 substrate has the rutile (tetragonal) lattice structure as the metallic phase of VO2. For
the rest of this section, most of the considerations are referred to as the rutile phase of VO2.
TiO2 substrate is oriented along the (001) surface, thus for the lattice mismatch calculation, the
a and b structural parameters have to be addressed.
The in-plane lattice mismatch between sample and substrate can be calculated as: 𝑀 =𝑎𝑠−𝑎𝑓
𝑎𝑓 ∗ 100 (2.1.1)
Where as and af are the lengths of cell edges of the substrate and of the sample, respectively.
Between rutile TiO2 (a=b=4.58 Å) and bulk VO2 (a=b=4.55 Å) there is a lattice mismatch M=0.66%.
Since titanium dioxide has larger unit cell edges, during the early stage of epitaxial grow, VO2
film will undergo a tensile strain in order to match the substrate lattice. For a thin film, the net effect is that of an increase of ar and br lattice constant and the consequent elastic compression
of cr [24,62,64]. Increasing the overlayer thickness, the influence of the substrate fades and the
lattice constants relax to the bulk value.
For the samples analyzed in this thesis, the critical thickness above which the sample can be considered bulk-like is 25.5 nm [64].
Figure 2.1.2: Schematic representation of the rutile metallic VO2 unit cell (left) and oxygen octahedron surrounding
vanadium atoms (right) for bulk and strained VO2/TiO2(001). The oxygen octahedra are depicted to underline the
orthorhombic distortion that allows differentiating between equatorial and apical oxygen. The mismatch between TiO2 (ar=br=4.58 Å) and VO2 (ar=br=4.55 Å) increases the ar and br lattice parameter in the epitaxial film and thus
decrease cr. This results in an increase in the apical V-O bond length.
From equations 1.2.1-2 we can observe that Ja increases as ar, i.e. the strain, increases, while Je
22
The increase of the apical bond length reduces the overlap between oxygen and vanadium orbitals and therefore the 3d-2p hybridization. The π* orbital is the most affected. The decrease in metal-ligand hybridization results in a decrease of the bonding-antibonding energy separation, consequently the π* orbital is downshifted in energy.
The 𝑑||∗ experience the opposite situation, decreasing cr increases the orbital overlap between
vanadium atoms, upshifting the 𝑑||∗.
Figure 2.1.3: Orbital evolution as a function of strain for the metallic (top panel) and insulating phase (bottom panel).
In the metal phase the empty part of the 𝑑|| band is referred to as 𝑑||∗.
The control over the lattice degree of freedom is critical in order to tune the electronic properties of VO2.
A detailed overview of the interplay between lattice and electronic structure is exposed in Chapters 3 and 4 where three epitaxial films of VO2/TiO2(001) with thickness 8, 16 and 32 nm
23
2.2 Nanostructured VO
xsamples synthesis: Supersonic
Cluster Beam Deposition
The class of NS includes solids whose constitutive elements have dimensions ranging from few, up to tens of nm. This definition stands for materials made of different building blocks e.g. nano-crystalline domains, nanoparticles or clusters, nm thin heterostructures, etc. In literature there is not a sharp difference between nanoparticles and clusters; however, there is a general tendency to consider clusters as a small aggregate of few atoms, while nanoparticles indicate an aggregate of at least 3*102 atoms. In this thesis clusters and nanoparticles will be used as
synonyms.
NS materials of great interest are those assembled by atomic clusters. These are an aggregation of atoms standing at midway between free atoms and bulk solids. The reduced dimensions guarantee a higher surface to volume (S/V) ratio respect to the bulk, increasing their chemical reactivity [69]. Moreover, their optical, structural and electronic properties depend on the number of constituent atoms [70–72]. In principle, precise control of the synthesis and knowledge of the aggregation properties could lead to the production of nanostructured materials, using clusters as “super-atoms”.
Accurate control over the synthesis parameter is particularly important for VOx NS films due to
the huge variety of oxidation state of vanadium [73].
Supersonic Cluster Beam Deposition (SCBD) has emerged as an invaluable tool for the synthesis of cluster assembled materials because of its high stability and intense cluster production [74]. In the next section the general principles of supersonic expansion, on which SCBD relies, will be presented.
2.2.1 Supersonic expansion of cluster beam
The basic thermodynamic considerations reported in this section have the purpose to highlight the main features of the supersonic expansion of the cluster beam. A complete thermodynamic description can be found in the literature [75,76].
A cluster beam, or more in general, a molecular beam, originates from the expansion of a gas from a high-pressure region P0 to a low-pressure region Pb (usually high vacuum). The expansion
between the two regions occurs through a low conductance nozzle. If P0/Pb is high enough the
24
Figure 2.2.1: Schematic representation of an apparatus for molecular beam productions.
The produced beam is extracted from the expansion chamber and brought to the experimental chamber using a conical skimmer before the turbulence region induced by the Mach disc. In general, if we consider a continuous flow of an ideal gas, the supersonic expansion can be considered isentropic neglecting viscosity and heat transfer effects. It is convenient for studying the system in terms of enthalpy. The stagnation enthalpy of the source chamber can be written as:
𝐻0 = 𝑐𝑝𝑇0 (2.2.1)
Where H0 is the stagnation enthalpy, cp is the constant pressure molar specific heat and T0 the
initial temperature of the source. During the expansion because of energy conservation, part of the stagnation enthalpy is converted into kinetic energy.
𝐻0 = 𝐻 + 1 2𝑚𝑣 2 = 𝑐 𝑝𝑇 + 1 2𝑚𝑣 2 (2.2.2)
In isentropic approximation, a supersonic expansion implies a cooling of the gas jet. In fact, equation 2.2.2 is valid only if T0>T, therefore part of the energy of the other degrees of freedom
converts into translational energy.
If the expansion is efficient the residual enthalpy H tends to 0 (T<<T0) therefore the limit value
of the kinetic energy is
1 2𝑚𝑣𝑡
2= 𝑐
𝑝𝑇0 (2.2.3)
Where vt is the terminal velocity. It is useful to remind that for an ideal gas, the following
relations are valid:
𝑐𝑣=12𝑓𝑅 (2.2.4) 𝑐𝑝 = 𝑐𝑣+ 𝑅 (2.2.5) 𝑐𝑝 𝑐𝑣= 𝛾 (2.2.6) 𝑐𝑝 = 𝑅𝛾 𝛾−1 (2.2.7)
Where cv is the constant volume specific heat, cp is the constant pressure specific heat, R is the
25 𝑣𝑡 = √ 2 𝑚𝑇0𝑅(1 + 1 2𝑓) = √ 2 𝑚𝑇0𝑅 𝛾 𝛾−1 (2.2.8)
Supersonic beams can also be pulsed i.e. the source is turned on for a short interval
periodically in time. The main advantage of using a pulsed source is the peak intensity is higher respect to continuous sources. In fact, Pb is considerably lower since the background gas is
pumped away while the source is turned off, allowing to gain up to a factor of 103 on the peak
intensity.
The thermodynamic relations described in this section hold also in the case of small perturbation of the “ideal gas” hypothesis. When the molecular beam contains particles of different species, it is called a seeded beam. In SCBD an inert gas (or carrier gas) is seeded with a diluted quantity of clusters (1-10% of the total pressure) so that the seeded beam can be treated as a small perturbation of the pure gas case. The cluster beam is composed of different species (the carrier gas atoms and clusters with different masses). Therefore the specific heat, the terminal velocity, and the average mass need to be defined using a weighted average. Each element of mass mi has energy:
𝐸𝑖≈ 𝑚𝑖
𝑚̅ 𝑇0 (2.2.9)
Where 𝑚̅ is the average mass and mi the mass of the i-th specie. In synthesis, a heavy species
will be accelerated if diluted into a light carrier gas and vice versa. Taking into account the difference in mass between the carrier gas and the clusters is of fundamental importance for the definition of the NS film properties as will be explained in the next section.
2.2.2 Pulsed Micro-Plasma Cluster Source
Pulsed Micro-Plasma Cluster Source (PMCS) [74,77,78] is a pulsed cluster source which proved its importance in nanostructured materials synthesis [79–81] and clusters fundamental study [82,83].
26
A schematic representation of the PMCS hardware is reported in Figure 2.2.2. A solenoid pulsed valve overlooks within the thermalization cavity. An electrode (anode) separates the ceramic body and the pulsed valve. The rotating cathode is inserted in a 6 mm hole that crosses the ceramic body perpendicularly to the cluster beam propagation direction. The PMCS works as follows: the pulsed valve injects the carrier gas within the ceramic body with an aperture of the order of hundreds of microseconds creating a region of higher gas density between the anode and the cathode. A pulsed potential difference (500-1000 V) is applied to the electrodes, with a delay of hundreds of microsecond respect to the valve aperture, triggering an intense electrical discharge (about 1000 A) with a duration of tens of microseconds. Consequently, between the electrodes, a plasma of ionized gas is generated, which will impinge the cathode ablating atoms from its surface. The cavity is now filled with an oversaturated gas mixture (carrier gas + ablated atoms) that undergoes a thermalization process starting clusters nucleation and condensation during the supersonic expansion of the beam. The aerodynamic lenses are necessary to improve clusters focusing (especially the small clusters) along the beam direction of propagation and, in addition, using the aerodynamic lenses increases the deposition rate [77,84].
One of the PMCS strengths is the ability to synthesize NS film minimizing clusters coalescence effects. To understand how it is possible, it is useful to study the cluster formation process and the critical parameters to take into account during the deposition stage.
The clusters formation in a seeded beam can be seen as the first stage of a phase transition, in particular, gas condensation. A complete theoretical description of the nucleation process [85– 88] goes beyond the purpose of this thesis, however, a qualitative description of the classical nucleation theories is helpful to understand how clusters are generated.
In classical nucleation theory, the clusters’ formation is treated as a gas-liquid phase transition [88]. The fundamental ansatz is to assume that the properties of the clusters can be extrapolated from those of a macroscopic quantity of the liquid phase, independently of cluster size. Therefore, a value of surface tension and chemical potential are assigned to the cluster, as it was a spherical droplet. The Gibbs free energy, i.e. the minimum work required to cluster formation has the qualitative trend shown in Figure 2.2.3.
27
The value n* is a critical point of the Gibbs free energy and defines two size regions. For n<n* the clusters grow increasing their free energy, this region is called embryos region. When n>n*the clusters growth is favoured by the decreasing of the free energy. Nucleation is a process in which the embryos clusters need to overcome an energy barrier defined by G*(n*) in order to grow until observable dimensions. To condensate the gas need to be oversaturated, i.e. for a given temperature, the pressure P must be higher than the vapour pressure Pv. In this model,
the processes responsible for cluster formation are monomer addition and evaporation, which compete to reach a steady-state where the growth rate is constant.
This classical viewpoint is oversimplified and it cannot properly describe the process of cluster formation inside a seeded beam. To improve the model, it is necessary to consider that the clusters are diluted in the carrier gas and that the monomer addition cannot be the only growth channel available [89]. Nevertheless, the general considerations over free energy of classical nucleation theory are applicable also to supersonic cluster sources. In a supersonic source, during the first stages of expansion, the gas-cluster mixture is cooled down by particles’ collisions. When a cluster reaches the size of n* the growth is then favoured by low energy monomer-cluster and cluster-cluster collisions. The mass distribution of the cluster is determined by the efficiency of these collisions, the geometry of the cluster source and the operational parameters [75].
After the nucleation process, the clusters beam is deposited on a substrate. The different deposition conditions are of paramount importance to determine the nanostructured material properties. The crucial parameter is the relative magnitude of the kinetic energy per atom E and the binding energy of an atom within the cluster ε. Upon landing, if E> ε the cluster will fragment and a strong coalescence is expected, while if E<ε clusters survive to the impact maintaining their individuality. In supersonic cluster sources usually, the value of E is in the range 0.1-1 eV/atom and considering the average binding energy of metal clusters [89], SCBD emerges as a low energy deposition technique. The different influence of cluster impact energy on nanostructured materials property has been widely studied in the literature using experimental and theoretical methods [90–93]. It has been shown that during the deposition of clusters with a kinetic energy per atom of the order of E≈1 eV fragmentation and coalescence may occur [91]. On the other hand, if E≈0.1 eV the impact with the substrate slightly perturb the cluster avoiding fragmentation and coalescence with other clusters. Therefore in order to synthesize NS samples that retain a memory of their gas-phase precursors, it is necessary to accurately control the kinetic energy at the moment of deposition.
Using equation 2.2.8 the terminal velocity of a cluster at the moment of deposition for a He seeded cluster beam and for an Ar seeded cluster beam can be estimated: vtHe=1766 m/s and
vtAr=588.6 m/s.
The kinetic energy per atom have been calculated for vanadium clusters: EHe=0.85 eV/atom
EAr=0.09 eV/atom
28
An additional strength of the PMCS is its extreme versatility: the ablation-by-discharge clusters synthesis is one of the few methods available to produce NS film of all kinds of metals, including refractory metals. The “classical” synthesis methods, like MBE, work mainly by evaporation of the metal of interest. This could constitute a technical problem when dealing with materials with a high melting point. In addition, often dedicated oven for different materials is necessary reducing the versatility of the evaporation methods. On the other hand, the only required condition to use a PMCS is to have a conductive cathode. Even semiconductors can be used as target materials, taking into account that heating of the cathode could be necessary.
Despite the enormous applicative potential, SCBD by mean of a PMCS has been mostly used to produce NS film of pure metal, carbon and transition metal oxide with the highest oxidation state [78,79,94,95]. The demonstration that NS film with controllable oxidation state can be produced during the synthesis and not just by air passivation, is still lacking.
During this thesis, we developed a procedure to oxidize the vanadium clusters during the early stage of their synthesis. To obtain homogeneously oxidized nanoparticles, we used Ar (high purity Ar: 99.9995%, SIAD) as carrier gas, mixed with a controlled amount of oxygen (Table 2.1) resulting in an Ar-O2 gas mixture. To probe the oxidation state of the samples, we used X-ray
Photoelectron Spectroscopy (XPS) as described in section 2.2.3.
The samples from section 2.2.3 to 2.2.6 have been synthesized in-situ under ultra-high vacuum (UHV) conditions (base pressure <210-9 mbar) by using the SCBD apparatus equipped with a PMCS which is available at Laboratory TASC-Analytical Division [96]. The PMCS was operated with a vanadium cathode (6 mm diam. rod, purity 99.9 %, EvoChem GmbH) generating a supersonic beam of vanadium metal or oxide cluster. The working parameters of the PMCS have been kept constant for all the samples synthesized (delay between gas injection and discharge firing = 0.6 ms; discharge operating voltage 0.925 kV; discharge duration 80 µs; pulsed-valve aperture driving signal duration time 157 µs; pulse repetition rate 3 Hz; Ar pressure 70 bar). The nanostructured film deposition rate measured by a quartz-crystal microbalance ranged from ~30 Å/s for pure Ar as the carrier gas, to ~5 Å /s for the highest oxygen concentration.
To probe the vanadium oxidation state, we deposited VOx films with >300 nm thickness
(measured using a quartz microbalance) on Si or Cu substrates. The typical size of all the samples is about 2 cm2.
Table 2.1: Stoichiometry of the NS films synthesized at different oxygen partial pressure in the Ar-O2 carrier gas
mixture. Oxygen partial pressure [mbar] x in VOx 15 ± 5 0.6 ± 0.02 40 ± 5 1.5 ± 0.02 50 ± 5 1.6 ± 0.02 60 ± 5 2.2 ± 0.02 70 ± 5 2.45 ± 0.02
29
Photoemission Spectroscopy (XPS, section 2.2.3) using an Mg Kα lamp (1253.6 eV, not monochromatic) and studied by UPS using a He lamp (UPS, section 2.2.5, He I 21.22 eV) coupled to a hemispherical electron analyzer (PSP, 120 mm). The morphology and the size of the clusters have been analyzed by Transmission Electron Microscopy (TEM) by means of a field-emission TEM (JEM2100F-UHR, JEOL) operated with beam energy 200 KV.
2.2.3 Stoichiometry determination: XPS
Technological applications require precise control over materials properties. Synthesizing samples with the necessary features to match the desired applications is of enormous importance. Vanadium oxides exhibit a wide range of stoichiometric (VO, V2O3, VO2 and V2O5)
and mixed-valence oxides (Magneli and Wadsley series) [6]. The ability to select a specific VOx
stoichiometry during sample synthesis is crucial to exploit vanadium oxides' applicative potential.
To investigate the stoichiometry, i.e., the ratio between oxygen and vanadium atoms, of the produced samples, we performed core-level XPS spectra of V 2p and O 1s electrons. A complete overview of XPS working principles is reported in Appendix 1. Quantitative information has been extracted by fitting simultaneously the line-shape of V 2p and O 1s core-levels which is a known indicator of the stoichiometry [97]. For each component (vanadium and oxygen) we used a pseudo-Voigt curve [98]. The ratio of the areas of V and O contributions, multiplied by the appropriate factors (cross-section, detector sensitivity and spin-orbit branching ration) result in the stoichiometric ratio x= [# of oxygen atoms]/[# of vanadium atoms]. From now on we use the stoichiometric ratio thus measured, in the range 0≤x≤2.5 to identify VOx samples since exists a
one to one correspondence between x and a specific sample.
Figure 2.2.4: V 2p and O 1s spectra of vanadium and oxygen. a) V 2p and O 1s core-level spectra of all samples. Peak positions of the vanadium components for different oxidation states are highlighted by straight vertical lines; b)
V 2p and O 1s core-level spectra and fit curve of VO2.2. The individual fit components are shown as coloured curves.
The label O 1s sat indicates the features generated by the Mg Kα3 and Mg Kα4 component of the radiation.
30
proves that the applied method may be used to synthesize VOx films with controllable
stoichiometry in the range 0.5<x<2.45.
Films with stoichiometry x<1 (VO0.53 in Figure 2.2.4) are called sub-oxides, i.e., oxides in which
the electropositive element is more abundant than oxygen. Generally, only a severe temperature treatment (T>1000 K) under high vacuum allows obtaining vanadium sub-oxides [99,100] while using the SCBD method we are able to synthesize them at room temperature.
2.2.4 3d occupancy investigation: Auger L
3M
23M
45spectroscopy
The stoichiometric vanadium oxides: VO, V2O3, VO2 and V2O5 have nominally 3, 2, 1 and 0 3d
unpaired electrons, respectively. Auger decays with one hole in the VB (i.e., of the form V XYM4,5)
probe the V 3d partial density of states (DOS) and therefore can be linked to 3d occupation number and to sample stoichiometry. The ideal Auger decay to probe the partial 3d DOS is the V L3M45M45,characterized by two holes in the valence band. Unfortunately, in vanadium oxides,
this decay channel is obscured by the O K L23L23 Auger electrons, even for an extremely low
amount of oxygen [101]. V L3M23M45 Auger decay is experimentally observable and 3d electrons
give the main contribution to this channel [102,103].
Figure 2.2.5: Comparison of Auger L3M23M45 spectra of vanadium oxides NS films with stoichiometric ratio 1.27, 1.6
and 2.2 in the kinetic energy range 455-482 eV. The spectra are vertically shifted for clarity shifted in order to highlight the behaviour of features labelled U and V (see text).
As shown in Figure 2.2.5 the Auger L3M23M45 of vanadium oxides exhibits two main features
labelled V and U. As pointed out by Sawatzky and Post, these two components are correlated with the oxidation state[104]. Feature V becomes dominant as the oxidation state increases: it is related to V 3d electrons covalently bound to O 2p electrons. Feature U is associated with unpaired 3d electrons. To extract quantitative information about the 3d contribution in the VB, we fit the Auger V L3M23M45 spectra using three pseudo-Voigt curves: one for feature V, one for
feature U, and the last one for the small contribution from the O K L1L1 decay at ~477 eV. The
31
and V refer to the area of the two features), can be associated with the fraction of unpaired 3d electrons relative to the total number of 3d electrons. The behaviour of the Auger fraction vs. 𝑥 is reported in Fig. 2.2.6 along with the best fit using a straight line.
Figure 2.2.6: Correlation between stoichiometry and Auger fraction (𝑈+𝑉𝑈 ) as obtained by fitting the Auger L3M23M45
spectra. The blue line is the best least-squares fit of the experimental data with a straight line of fixed intercept (at x=0 the fraction of unpaired electrons must be equal to 1). The best-fit slope is k = -0.3200.005
In the pure metal, no oxygen atoms are available to form V-O bonds, thus V = 0 and U/(U+V)=1; accordingly, in the straight-line fit, the intercept is held at 1. Likewise, in the maximum oxidation state (x=2.5) the fraction should be zero because no unpaired 3d electrons are present and U=0. Actually, the extrapolated value for x=2.5 is different from zero: U/(U+V)|2.5 =0.2±0.01, denoting
a partial 3d occupancy. This is not unusual for 3d0 compounds because of the strong 3d-2p
hybridization in the specific case of V2O5, these results confirm previous published resonant
photoemission (ResPES) experiments [105,106]. In Table 2.2 we report the number of unpaired
3d electrons per vanadium atom equal to 3U/(U+V) under the above assumptions. They are not
in good agreement with the occupation number calculated by Zimmerman and co-workers [107], suggesting that further theoretical investigations are necessary.
We would also like to underline here that using Auger spectroscopy, the investigated system, is the core-hole ionized system, and do not necessarily reflect neutral VOx features. For example,
while the extrapolated values are in good agreement with measured values of unpaired 3d electrons for VO2, suggesting that the core-hole effect in this oxide is negligible. On the other
hand, the data for the V2O3 oxide suggest that only 1.56, instead of 2, 3d electrons are unpaired.
